Support dynamic error sampling

### Description

Just like `TracesSamplerCallback`, we want to have a callback capable sampling errors. This feature already exists in our [Python SDK](https://docs.sentry.io/platforms/python/configuration/sampling/#dynamically-sampling-error-events).

```Python
sentry_sdk.init(
# ...

error_sampler=my_sampler,
)
```

Right now the workaround would be to keep hold of the options and change the `sampleRate` after the SDK has been initialized.
This should work just fine, as `SentryClient` reads out the sample rate every time, but is still discouraged, as we don't officially support changing options on the fly.

```Java
final AtomicReference optionsRef = new AtomicReference<>();
Sentry.init(new Sentry.OptionsConfiguration() {
@Override
public void configure(@NotNull SentryOptions options) {
options.setDsn("...");
options.setSampleRate(1.0); // some sane default
// ...
optionsRef.set(options);
}
});

// later, dynamically update the sampleRate
private void onRemoteConfigChanged(final @NotNull RemoteConfig remoteConfig){
final @Nullable SentryOptions options = optionsRef.get();
if (options != null) {
options.setSampleRate(remoteConfig.sampleRate);
}
}
```

先閱讀現有的 TracesSamplerCallback、SentryClient 和 SentryOptions 流程，尤其是如何為錯誤讀取 sampleRate。在現有的 trace sampler 旁定義錯誤採樣回呼，並驗證錯誤事件可以動態採樣，而不必在初始化後變更選項。當 Java SDK 支援所要求的回呼行為，且測試涵蓋採樣結果時，即表示完成。

由索引模型根據 Issue 內容生成。